bug#44399: Cuirass problem with building from manifest file

bug#44399: Cuirass problem with building from manifest file

[david larsson]

Hi!

I am unable to build a simple manifest containing the hello package, see 
below. I tried imitating the manual 
https://guix.gnu.org/manual/en/html_node/Continuous-Integration.html but 
perhaps the manifest.scm is not supposed to contain a regular manifest 
or maybe that config example is outdated?

============================
The hello.scm file - works fine using guix package -m hello.scm
============================

cat /home/cuirass/my-guixsd-config.sh/manifests/hello.scm
(specifications->manifest
  (append (list "hello")
          '()))

============================
 From /var/log/cuirass.log
============================

2020-11-02T20:06:25 fetching input 'my-guixsd-config' of spec 'my-pkgs'
2020-11-02T20:06:25 fetching input 'my-guix-packages' of spec 'my-pkgs'
2020-11-02T20:06:25 fetching input 'guix' of spec 'my-pkgs'
2020-11-02T20:06:26 fetched input 'my-guixsd-config' of spec 'my-pkgs' 
(commit "fb3001bd453583b4638bf6ec7187b5758d7b0715")
2020-11-02T20:06:26 fetched input 'my-guix-packages' of spec 'my-pkgs' 
(commit "9f9c644076b088aabee76626598115c0fe6bfbde")
2020-11-02T20:06:26 fetched input 'guix' of spec 'my-pkgs' (commit 
"7c02eb6c8d7d050ecb8f3082c4a754ad32211403")
2020-11-02T20:06:26 next evaluation in 36000 seconds
2020-11-02T20:06:26 evaluating spec 'my-pkgs'
Uncaught exception in fiber ##f:
In cuirass/base.scm:
    785:13  3 (_)
In ice-9/boot-9.scm:
   1736:10  2 (with-exception-handler _ _ #:unwind? _ # _)
   1669:16  1 (raise-exception _ #:continuable? _)
   1667:16  0 (raise-exception _ #:continuable? _)
ice-9/boot-9.scm:1667:16: In procedure raise-exception:
ERROR:
   1. &evaluation-error:
       name: "my-pkgs"
       id: 1

============================
My cuirass-specs
============================

(define %cuirass-specs
   #~(
      list
   '((#:name . "my-pkgs")
     (#:load-path-inputs . ("guix"))
     (#:package-path-inputs . ("my-guix-packages"))
     (#:proc-input . "guix")
     (#:proc-file . "build-aux/cuirass/gnu-system.scm")
     (#:proc . cuirass-jobs)
     (#:proc-args .
      ((subset . manifests)
       ;; This builds fine!
       ;;(subset . ("hello" "cowsay" "orgmk" "fwknop" "python-on-guile" 
"libxml2-xpath0" "guile-bash-parallel" "guile-base64" "perl-data-ical" 
"python-pydotplus"))
       (systems . ("x86_64-linux"))
        ;; This fails on 7c02eb6
       (manifests . (("my-guixsd-config" . "hello.scm")))
       ))
     (#:inputs . (
                  ((#:name . "guix")
                   (#:url . "git://git.savannah.gnu.org/guix.git")
                   (#:load-path . ".")
                   ;;(#:branch . "master")
                   (#:commit . 
"7c02eb6c8d7d050ecb8f3082c4a754ad32211403")

                  ((#:name . "my-guix-packages")
                   (#:url . "file:///home/cuirass/my-guix-packages")
                   (#:load-path . "packages")
                   (#:branch . "master")
                   (#:no-compile? . #t))

                  ((#:name . "my-guixsd-config")
                   (#:url . "file:///home/cuirass/my-guixsd-config.sh")
                   (#:load-path . "manifests")
                   (#:branch . "master")
                   (#:no-compile? . #t))

                  ))
     (#:build-outputs . ()))
))

============================
And the evaluations file content
============================

gzip --decompress -c /var/log/cuirass/evaluations/1.gz
warning:
building things during evaluation
'build-things' arguments: 
(("/gnu/store/qhm7xsikyykvjisi8kbmzp8ny3cmf370-compute-guix-derivation.drv") 
0)
Computing Guix derivation for 'x86_64-linux'...
warning:
building things during evaluation
'build-things' arguments: 
(("/gnu/store/ps30mn7xq0ymcrdzlwim8pkv1czgv2zf-profile.drv") 0)
ERROR: In procedure read:
In procedure scm_lreadr: #<unknown port>:16:634: Unknown # object: #\<



Best regards,
David Larsson

bug#44399: Cuirass problem with building from manifest file

[Mathieu Othacehe]

Hello David,

> (("/gnu/store/ps30mn7xq0ymcrdzlwim8pkv1czgv2zf-profile.drv") 0)
> ERROR: In procedure read:
> In procedure scm_lreadr: #<unknown port>:16:634: Unknown # object: #\<

This is probably a consequence of me breaking things today. I restored
the original handling of package licenses with b5f2a035.

This means that you need to uses non "#f" licenses in your custom
packages to avoid the issue you encountered in
https://issues.guix.gnu.org/44370.

Could you please confirm that it works?

Thanks,

Mathieu

bug#44399: Cuirass problem with building from manifest file

[david larsson]

On 2020-11-02 20:12, Mathieu Othacehe wrote:
> 
> Could you please confirm that it works?
> 

It does not. Building from manifest gives no error at all but goes 
stale. The last output line in /var/log/cuirass.log is just:

2020-11-03T07:57:46 evaluating spec 'my-pkgs'

htop shows no cpu usage.



Building packages with #f licenses throws the same error again as in 
#44370.



Packages are otherwise built but reported as failed, as in 
https://issues.guix.gnu.org/44390

bug#44399: Cuirass problem with building from manifest file

[david larsson]

On 2020-11-02 20:12, Mathieu Othacehe wrote:

> 
> Could you please confirm that it works?
> 
> Thanks,
> 
> Mathieu


Things do work now. I had made a mistake in my config.

In case it helps someone; to fix it, I changed my-guixsd-config input 
load path to ".":

                  ((#:name . "config")
                   (#:url . "file:///home/cuirass/my-guixsd-config.sh")
                   (#:load-path . ".")
                   (#:branch . "master")
                   (#:no-compile? . #t))

and in proc-args I changed the manifests path, and made sure "manifests" 
has quotes around it, so I changed it to:

     (#:proc-args .
      ((subset . "manifests")
       (systems . ("x86_64-linux"))
       (manifests . (("config" . "manifests/user1.scm")))
       ))

which now built over 200 packages successfully, yay!


Thanks for your help Mathieu!

Best regards,
David

bug#44399: Cuirass problem with building from manifest file

[Mathieu Othacehe]

Hello,

As the specification format is completely new and the evaluation process
has been rewritten, I think we can close this one.

Thanks,

Mathieu